Skip to content

Cleaning a PostgreSQL Database

In this project, you will work with data from a hypothetical Super Store to challenge and enhance your SQL skills in data cleaning. This project will engage you in identifying top categories based on the highest profit margins and detecting missing values, utilizing your comprehensive knowledge of SQL concepts.

Data Dictionary:

orders:

ColumnDefinitionData typeComments
row_idUnique Record IDINTEGER
order_idIdentifier for each order in tableTEXTConnects to order_id in returned_orders table
order_dateDate when order was placedTEXT
marketMarket order_id belongs toTEXT
regionRegion Customer belongs toTEXTConnects to region in people table
product_idIdentifier of Product boughtTEXTConnects to product_id in products table
salesTotal Sales Amount for the Line ItemDOUBLE PRECISION
quantityTotal Quantity for the Line ItemDOUBLE PRECISION
discountDiscount applied for the Line ItemDOUBLE PRECISION
profitTotal Profit earned on the Line ItemDOUBLE PRECISION

returned_orders:

ColumnDefinitionData type
returnedYes values for Order / Line Item ReturnedTEXT
order_idIdentifier for each order in tableTEXT
marketMarket order_id belongs toTEXT

people:

ColumnDefinitionData type
personName of Salesperson credited with OrderTEXT
regionRegion Salesperson in operating inTEXT

products:

ColumnDefinitionData type
product_idUnique Identifier for the ProductTEXT
categoryCategory Product belongs toTEXT
sub_categorySub Category Product belongs toTEXT
product_nameDetailed Name of the ProductTEXT

As you can see in the Data Dictionary above, date fields have been written to the orders table as TEXT and numeric fields like sales, profit, etc. have been written to the orders table as Double Precision. You will need to take care of these types in some of the queries. This project is an excellent opportunity to apply your SQL skills in a practical setting and gain valuable experience in data cleaning and analysis. Good luck, and happy querying!

Spinner
DataFrameas
top_five_products_each_category
variable
-- top_five_products_each_category
WITH product_totals AS(
	SELECT p.category,
		p.product_name,
		SUM(o.sales) AS product_total_sales_raw,
		SUM(o.profit) AS product_total_profit_raw
	FROM public.orders o
	JOIN public.products p
	    ON o.product_id = p.product_id
	GROUP BY p.category, p.product_name
), ranked AS (
	SELECT	
	    category, 
	 	product_name,
	 	ROUND(CAST(product_total_sales_raw AS NUMERIC), 2)  AS product_total_sales,
		ROUND(CAST(product_total_profit_raw AS NUMERIC), 2) AS product_total_profit,
	 	RANK () OVER (
	 	 	PARTITION BY category ORDER BY product_total_sales_raw DESC
		) AS product_rank
	FROM product_totals
)
SELECT 
 	category,
	product_name,
	product_total_sales,
	product_total_profit,
	product_rank
FROM ranked
WHERE product_rank <= 5
ORDER BY category ASC, product_total_sales DESC, product_name ASC;
Spinner
DataFrameas
impute_missing_values
variable
WITH
-- 1) Rows where quantity is missing
missing AS (
  SELECT
    o.product_id,
    o.discount,
    o.market,
    o.region,
    o.sales,
    o.quantity
  FROM public.orders o
  WHERE o.quantity IS NULL
),

-- 2) Unit price per product_id + discount from available data
unit_prices AS (
  SELECT
    o.product_id,
    o.discount,
    CAST(SUM(o.sales) / NULLIF(SUM(o.quantity), 0) AS NUMERIC) AS unit_price
  FROM public.orders o
  WHERE o.quantity IS NOT NULL AND o.quantity > 0
  GROUP BY o.product_id, o.discount
)

-- Final: compute calculated_quantity for missing rows
SELECT DISTINCT
  m.product_id,
  m.discount,
  m.market,
  m.region,
  m.sales,
  m.quantity,
  ROUND( CAST(m.sales AS NUMERIC) / up.unit_price, 0 ) AS calculated_quantity
FROM missing m
INNER JOIN unit_prices up
  ON up.product_id = m.product_id
 AND up.discount  IS NOT DISTINCT FROM m.discount;